Validity of environmental audits using GigaPan® and Google Earth Technology

Abstract

Health behaviors are shaped by the context in which people live. However, documenting environmental context has remained a challenge. More specifically, direct observation techniques require large investments in time and resources and auditing the environment through web-based platforms has limited stability in spatio-temporal imagery. This study examined the validity of a new methodology, using GigaPan ® imagery, where we took photos locally and, stitched them together using GigaPan ® technology, and quantified environmental attributes from the resulting panoramic photo. For comparison, we examined validity using Google Earth imagery. A total of 464 street segments were assessed using three methods: GigaPan ® audits, Google Earth audits, and direct observation audits. Thirty-seven different attributes were captured representing three broad constructs: land use, traffic and safety, and amenities. Sensitivity (i.e. the proportion of true positives) and specificity (i.e. the proportion of true negatives) were used to estimate the validity of GigaPan ® and Google Earth audits using direct observation audits as the gold standard. Using GigaPan ® , sensitivity was 80% or higher for 6 of 37 items and specificity was 80% or higher for 31 of 37 items. Using Google Earth, sensitivity was 80% or higher for 8 of 37 items and specificity was 80% or higher for 30 of 37 items. The validity of GigaPan ® and Google Earth was similar, with significant differences in sensitivity and specificity for 7 items and 2 items, respectively. GigaPan ® performed well, especially when identifying features absent from the environment. A major strength of the GigaPan ® technology is its ability to be implemented quickly in the field relative to direct observation. GigaPan ® is a method to consider as an alternative to direct observation when temporality is prioritized or Google Earth imagery is unavailable.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 11 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.